Articles https://doi.org/10.1038/s41589-018-0120-6

O-GlcNAc modification of eIF4GI acts as a translational switch in heat shock response

Xingqian Zhang1, Xin Erica Shu1,2 and Shu-Bing Qian 1,2*

Heat shock response (HSR) is an ancient signaling pathway leading to thermoprotection of nearly all living organisms. Emerging evidence suggests that intracellular O-linked β-​ N-acetylglucosamine (O-GlcNAc) serves as a molecular ‘thermometer’ by reporting ambient temperature fluctuations. Whether and how O-GlcNAc modification regulates HSR remains unclear. Here we report that, upon heat shock stress, the key translation initiation factor eIF4GI undergoes dynamic O-GlcNAcylation at the N-terminal region. Without O-GlcNAc modification, the preferential translation of stress mRNAs is impaired. Unexpectedly, stress mRNAs are entrapped within stress granules (SGs) that are no longer dissolved during stress recovery. Mechanistically, we show that stress-induced eIF4GI O-GlcNAcylation repels poly(A)-binding protein 1 and promotes SG disassembly, thereby licensing stress mRNAs for selective translation. Using various eIF4GI mutants created by CRISPR/Cas9, we demonstrate that eIF4GI acts as a translational switch via reversible O-GlcNAcylation. Our study reveals a central mechanism linking heat stress sensing, protein remodeling, SG dynamics and translational reprogramming.

ynamic glycosylation of proteins with O-GlcNAc has been These findings provide fundamental insights into SG regulation and rapidly emerging as a cellular signaling mechanism1–3. The generate new perspectives for translational control of HSR. Dreversible O-GlcNAcylation is mediated by a single pair of enzymes: O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA). Results While the former installs O-GlcNAc on serine or threonine resi- Deficient Hsp70 translation in cells lacking OGT. In mamma- dues of proteins4, the latter removes the monosaccharide unit from lian cells, O-GlcNAc modification has been implicated in HSR, proteins5. The dynamic cycling of O-GlcNAc modification occurs although the underlying mechanism remains to be resolved6,10. We in a nutrient- and stress-responsive manner. For many ectotherms, took advantage of a mouse embryonic fibroblast (MEF) cell line, O-GlcNAc levels are closely correlated with ambient temperature OgtF/Y(mER-Cre) MEFs, carrying an Ogt gene that can be excised during their development6. Likewise, heat shock stress rapidly by simple administration of 4-hydroxytamoxifen (4HT)10. As increases the levels of O-GlcNAcylation in mammalian cells7–9. expected, 48 h exposure to 4HT resulted in complete depletion of Notably, O-GlcNAc modification has been shown to confer heat OGT and disappearance of background O-GlcNAc (Fig. 1a). Upon stress resistance in endotherms such as humans10. However, neither heat shock stress, a robust induction of Hsp70 was evident in MEFs the identity of protein targets nor the physiological significance of before 4HT treatment. However, OGT depletion via 4HT pretreat- stress-induced O-GlcNAcylation has been clearly defined. ment led to marked attenuation of Hsp70 induction (Fig. 1a). The HSR is an evolutionarily conserved pathway that results in impairment in Hsp70 expression was not due to the side effect of rapid induction of genes encoding molecular chaperones essential 4HT, as the same treatment had little effect in control MEFs bearing for protection and recovery from cellular damages11–13. Upon heat GFP (Supplementary Fig. 1a). Additionally, OGT depletion did not shock stress, there is a marked inhibition of global protein synthesis yet lead to apoptosis or necrosis, at least under these experimental accompanied by the selective induction of stress proteins such as conditions (Supplementary Fig. 1b,c). The OGT-dependent Hsp70 Hsp70 (also termed Hsp72). This reprogramming of gene expres- induction also held true in HeLa cells, as RNA interference (RNAi)- sion involves a coordination of transcription14, translation15 and mediated OGT knockdown reduced Hsp70 levels (Supplementary post-transcriptional processes16,17. Additionally, a wide variety of Fig. 1d). The crucial role of OGT in HSR is not limited to Hsp70, stress stimuli trigger the assembly of SGs, discrete RNA–protein because small chaperones such as Hsp25 also showed substantial structures that contain nontranslating mRNAs18–21. SGs have been reduction in stressed MEFs lacking OGT (Supplementary Fig. 1e). proposed to affect mRNA translation and stability and have been However, constitutively expressed chaperones such as Hsc70 were linked to cellular processes such as apoptosis. Assembly and disas- minimally affected. sembly of SGs are modulated by various post-translational modi- Looking into the underlying mechanism, we found that OGT fications, as well as numerous protein-remodeling complexes21. depletion had negligible effect on Hsp70 mRNA levels in heat- However, very little is known about mRNA triage during SG for- stressed MEFs (Fig. 1a). To assess the translational status of Hsp70 mation. Whether there is a common mechanism underlying SG in these cells, we examined mRNA enrichment in polysome frac- dynamics and selective mRNA translation remains elusive. tions separated by sucrose gradient sedimentation22. The stress- In this study, we investigated the mechanistic linkage between induced Hsp70 mRNA, but not the housekeeping messenger B2m, O-GlcNAc modification and HSR in mammalian cells. We uncov- showed decreased polysome enrichment after OGT depletion ered a functional switch for eIF4GI that relies on O-GlcNAc modi- (Fig. 1b). These results suggest a crucial role for OGT in transla- fication to couple selective mRNA translation and SG dynamics. tional control of HSR.

1Division of Nutritional Sciences, Cornell University, Ithaca, NY, USA. 2Graduate Program in Nutritional Sciences, Cornell University, Ithaca, NY, USA. *e-mail: [email protected]

Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology 909 Articles NAtURE CHEMIcAl BIOlOGy

a 60 – 4HT b F /Y – 4HT Ogt (mER-Cre) MEFs + 4HT ** + 4HT – 4HT + 4HT 40 ** Post-HS N 00224466N (h)

20 OD 254 nm OGT Hsp70 protein 0 Fraction 24681012141618

250 Post-HS N 0 246 (h) 2.0 Hspa1a * 1.5 * 130 1.0 * 3,000 – 4HT

O -GlcNAc + 4HT 0.5 0.0 2,000 2.0 – 4HT 1.5 B2m Hsp70 + 4HT 1,000 1.0 Relative mRNA level s Hsp70 mRNA 0.5 β-actin 0 0.0 6810 12 14 Post-HS N 0246(h) Total 80S Polysome

Fig. 1 | Cells lacking OGT exhibit deficient Hsp70 translation. a, OgtF/Y(mER-Cre) MEFs were pretreated with (+​) or without (–) 0.5 µ​M 4HT for 24 h followed by normal medium for another 24 h. Cells were heat stressed (HS) at 42 °C for 1 h and recovered at 37 °C for indicated times. Whole-cell lysates were collected for immunoblotting and total mRNAs extracted for RT-qPCR. N: no heat shock. Top right: relative protein levels quantified by densitometry and normalized to β​-actin. Bottom right: relative levels of Hsp70 mRNA normalized to β​-actin. Basal values in wild-type cells are set to 1. Error bars, mean ±​ s.e.m.; **P <​ 0.01; two-way ANOVA; n = 3​ independent experiments; uncropped blots are shown in Supplementary Fig. 9. b, MEF cells as in a were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Whole-cell lysates were collected for polysome profile analysis as shown in the top panel. Total RNA was extracted from individual polysome fractions followed by RT-qPCR measuring β​2 microglobulin (B2m) and Hsp70 (Hspa1a) mRNA. Relative mRNA levels are normalized to the total. OD, optical density. Error bars, mean ±​ s.e.m.; *P <​ 0.05; two-way ANOVA; n =​ 3 independent experiments.

Heat shock stress induces O-GlcNAcylation of eIF4GI. We next 67.5%, MS revealed only one modification site with high confi- attempted to identify protein targets subjected to O-GlcNAc modi- dence, Ser68 (Fig. 2d). Notably, Ser68 in the mouse eIF4GI is well fication in response to heat shock stress. We initially focused on conserved in the human eIF4GI (annotated as Ser61). The Ser61 of ribosomal proteins, whose O-GlcNAcylation has been reported pre- human eIF4GI has previously been mapped as a putative modifica- viously23,24. However, translating ribosomes exhibited very low basal tion site26. To validate the identified modification site, we cloned levels of O-GlcNAc modification and showed negligible changes the full-length eIF4GI from HeLa cells and fused YFP and Flag to in response to heat shock stress (Supplementary Fig. 2a,b). We the N terminus. Stress-induced O-GlcNAc modification was well noticed that the most conspicuous O-GlcNAc species in ribosome preserved in transfected YFP-flag-eIF4GI (Fig. 2e). Introducing an fractions were of high molecular weight, and they disappeared in S68A mutation completely abolished O-GlcNAc signals from puri- OGT-depleted MEF cells after 4HT pretreatment (Supplementary fied eIF4GI fusion proteins (Fig. 2e). By contrast, S1285A mutation Fig. 2b,c). A close inspection of these high-molecular-weight spe- minimally affected the total O-GlcNAc levels of transfected eIF4GI. cies revealed a distinct band of approximate 250 kDa that emerged Therefore, the Ser68 is the primary site of eIF4GI that undergoes shortly after heat shock stress (Fig. 2a). Notably, the heat-stress- dynamic O-GlcNAcylation in response to heat shock stress. induced O-GlcNAc species appeared earlier than Hsp70 and van- ished during stress recovery. This dynamic feature was reproducible eIF4GI O-GlcNAcylation repels PABP1. The N-terminal third in HeLa cells (Fig. 2a), suggesting that the reversible O-GlcNAc of eIF4GI contains binding sites for poly(A)-binding protein 1 modification is a general phenomenon in mammalian cells upon (PABP1) and eIF4E27,28. We found that eIF4GI bound to eIF4E irre- heat shock stress. spective of O-GlcNAc modification (Fig. 3a). Notably, heat stress Given the putative role of O-GlcNAcylation in translational diminished the interaction between eIF4GI and PABP1 in wild- control of HSR, we surveyed several translation initiation factors type cells. However, in MEF cells lacking OGT, comparable amount that are relatively large in size. eIF4GI emerged as one of the candi- of PABP1 molecules were pulled down by eIF4GI before and after dates because of high scores of predicted O-GlcNAcylation and the stress (Fig. 3a). Given the close proximity of Ser68 and the PABP1 similar size of 250 kDa on gels (Supplementary Fig. 3a). To experi- binding site of eIF4GI, the stress-induced PABP1 dissociation is mentally validate this finding, we immunoprecipitated endogenous likely to be due to stress-induced O-GlcNAcylation of eIF4GI. eIF4GI from HeLa cells with or without heat shock stress. With sim- To test this possibility, we conducted reciprocal immunoprecipi- ilar input levels, eIF4GI obtained from heat-stressed cells showed a tation by purifying either eIF4E or PABP1 followed by eIF4GI marked increase in O-GlcNAc signal (Fig. 2b). We also examined detection. While eIF4E readily pulled down both modified and the eIF4GI paralog eIF4GII25, which appeared to be barely modified unmodified eIF4GI, much less eIF4GI was coprecipitated with in spite of the same size and sequence homology (Supplementary PABP1 and showed no O-GlcNAc signals (Fig. 3b). Notably, heat Fig. 3b). The stress-induced O-GlcNAc modification of eIF4GI shock stress also diminished the coprecipitation between PABP1 was further confirmed in MEF cells (Fig. 2c). Notably, OGT deple- and eIF4E. As an independent validation, we treated MEF cells tion after 4HT treatment completely abolished O-GlcNAc signals with thiamet G, an inhibitor of OGA. Consistent with the dynamic associated with eIF4GI. Therefore, heat shock stress triggers rapid nature of eIF4GI O-GlcNAcylation, we observed an increase of O-GlcNAc modification of eIF4GI. O-GlcNAc signals during the early stage of stress recovery (3 h) but We next applied tandem mass spectrometry (MS) to map the an evident decrease at the late stage (6 h) (Supplementary Fig. 3c). positions of O-GlcNAc modification on endogenous eIF4GI puri- In the presence of thiamet G, however, the O-GlcNAc modifi- fied from heat-stressed MEF cells. With a peptide coverage of cation of eIF4GI was sustained. Notably, much less PABP1 was

910 Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology NAtURE CHEMIcAl BIOlOGy Articles

a bc Ogt F /Y(mER-Cre) MEFs MEF HeLa HeLa eIF4GI IP Input N Post-HS N Post-HS eIF4GIIPInput Time 000011224466(h) HS – +–+ – 4HT + 4HT – 4HT + 4HT HS ––+ ++––+ 250 O-GlcNAc O-GlcNAc

O -GlcNAc 130 eIF4GI eIF4GI Hsp70

β-actin β-actin β-actin

[(z + 1)/Y ] d ×10 1 10 e EThcD 1,175.655 OH 100

O OH OH 80 Anti-Flag IP O NHAc eIF4GI YFP-flag-eIF4GI WT S68A S1285A 60 1,158.606 HS – ++––+ 4E 4A eIF3 4A

S68 [(z + 1)/Y1]5 679.432

O O-GlcNAc 2

b HexNAc 40 2 [(z + 1)/Y )] 200.014 1 9 1 500 1,000 1,500 2+ 1,087.583 [Y8/Y1] HexNAc·H 204.158 488.345 186.010 [(z + 1)/Y ] [(c – 1)/Y1)]9 Anti-Flag

Relative abundance 1 6 (z + 1)2 (z + 1)3 1,016.625 1,132.641 247.098 766.433 20 318.203 z 138.018 4 Y /Y Y /Y 388.179 7 1 8 1 MS coverage (67.5%) (z + 1) 2+ 5 [z /Y ] 878.598 975.510 476.299 10 1 Y /Y 956.584 587.518 5 1 694.432 0 200 400 600 800 1,000 m/z

Fig. 2 | eIF4GI undergoes dynamic O-GlcNAcylation in response to heat shock stress. a, OgtF/Y(mER-Cre) MEFs and HeLa cells were heat stressed (HS) at 42 °C for 1 h and recovered at 37 °C for the indicated times. Whole-cell lysates were collected for immunoblotting. N: no heat shock. Red arrowhead: putative O-GlcNAcylated eIF4GI. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. b, HeLa cells were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cell lysates were immunoprecipitated (IP) using an anti- eIF4GI antibody followed by immunoblotting. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. c, OgtF/Y(mER-Cre) MEFs were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cell lysates were immunoprecipitated using an anti-eIF4GI antibody followed by immunoblotting. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. d, Left: the schematic structure of eIF4GI with main domains color-coded for different protein interactions. Right: immunoprecipitated eIF4GI from heat-stressed MEF cells as in c was subjected to mass spectrometry. An electron transfer/higher energy collision dissociation (EThcD) MS/MS spectrum of m/z 587.79132+ identifies O-GlcNAcylation of Ser68 (red), illustrated along the peptide sequence shown at top. HexNAc, N-acetylhexoseamine. e, MEF cells were transfected with plasmids expressing YFP-flag-tagged eIF4GI wild type (WT) or mutants for 24 h. Transfected cells were heat stressed at 42 °C for 1 h and recovered at 37 °C for 4 h. Cell lysates were immunoprecipitated using an anti-Flag antibody followed by immunoblotting. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9.

­coprecipitated with eIF4GI (Supplementary Fig. 3c). Therefore, for the wild-type eIF4GI but not the S68A mutant (Supplementary stress-induced O-GlcNAc modification of eIF4GI remodels the Fig. 3d). In agreement with the endogenous eIF4GI results, a large assembled eIF4F complex by dissociating PABP1. amount of mRNA was significantly enriched in the S68A mutant in PABP1 is believed to bring the 5′ ​cap and 3′ ​end of mRNA comparison to the wild type (Fig. 3d). This unexpected result sug- into a closed loop and stimulate the cap-dependent translation gests that stress-induced O-GlcNAcylation of eIF4GI releases stress initiation29,30. It is conceivable that, by opening the mRNA loop, mRNAs that otherwise are trapped. O-GlcNAcylation of eIF4GI not only represses cap-dependent translation, but also facilitates cap-independent translation by eIF4GI O-GlcNAcylation promotes SG disassembly. We noticed remodeling the eIF4F complex. Since eIF4GI bears an inher- that, upon heat shock stress, less eIF4GI was recovered from the solu- ent mRNA binding capacity31, we speculated that stress-induced ble fraction in MEF cells lacking OGT (Fig. 3a, input). Fractionation O-GlcNAcylation led to preferential binding of eIF4GI to stress analysis further confirmed this feature (Fig. 4a). In addition to mRNAs for selective translation. To illustrate the mRNA-binding eIF4GI, eIF4E and PABP1 also reduced their solubility in OGT-null properties of eIF4GI with or without O-GlcNAc modification, we cells upon heat shock stress. This feature is reminiscent of stress enriched eIF4GI-associated mRNAs by zero-distance cross-linking­ granule (SG) formation in the cytoplasm32. Immunofluorescence followed by real-time quantitative qPCR (RT-qPCR) (Fig. 3c). indeed revealed more cytoplasmic puncta signified by SG markers in In control MEF cells, heat shock stress led to sixfold enrichment OGT knockout cells than in wild-type MEFs after heat shock stress of Hsp70 (Hspa1a) mRNA to eIF4GI, but not β​-actin (Actb) (Fig. 4b). In wild-type cells, heat-stress-induced SGs were transient mRNA. Surprisingly, and perhaps counterintuitively, OGT deple- and quickly resolved during stress recovery. In OGT knockout cells, tion resulted in about 25-fold enrichment of Hspa1a mRNA in however, the SG puncta largely remained even after the cessation association with eIF4GI (Fig. 3c). Since eIF4GI lacking O-GlcNAc of stress for 2 h (Supplementary Fig. 4a). To substantiate this find- modification does not support Hspa1a mRNA translation (Fig. 1), ing further, we knocked down OGT in HeLa cells and observed a the stable interaction between the unmodified eIF4GI and mRNA sustained formation of SGs during stress recovery (Supplementary is suggestive of nonproductive trapping. As an independent vali- Fig. 4b). Therefore, lack of O-GlcNAcylation results in delayed SG dation, we compared the mRNA-binding properties of YFP-flag- clearance during stress recovery. eIF4GI and the S68A mutant in transfected MEF cells. As expected, The unanticipated role of O-GlcNAc in SG disassembly is in sharp reduced PABP1 binding upon heat shock stress was only observed contrast to the results of a previous study that reported that OGT

Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology 911 Articles NAtURE CHEMIcAl BIOlOGy

a S68 b

IP eIF4GI PABP 4E 4A eIF3 4A eIF4E PABP1 Input HS ––++– + eIF4GI IP Input PAB 4G 4E 4HT ––++ – – ++ O-GlcNAc HS ––+ ++– – + O-GlcNAc eIF4GI HS eIF4GI

PABP1 PAB eIF4E

PABP1 eIF4E 4G 4E Hsp90 β-actin

c d 4SU

YFP YFP Y HS 40 25 Actb Actb Hspa1a 20 Hspa1a UV 365 nm 30 levels levels 15 20 10 10 5 Relative mRNA

4G Relative mRNA 0 0 HS – + – + YFPWT S68A 4HT ––++ qPCR YFP-flag-4GI

Fig. 3 | eIF4GI O-GlcNAcylation repels PABP1. a, OgtF/Y(mER-Cre) MEFs were heat stressed (HS) at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cell lysates were immunoprecipitated (IP) using an anti-eIF4GI antibody followed by immunoblotting. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. b, Left: HeLa cells were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cell lysates were immunoprecipitated using an anti-eIF4E or anti-PABP1 antibodies followed by immunoblotting. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. Right: a schematic of eIF4F complexes. 4E, eIF4E; 4G, eIF4GI; PAB, PABP1. c, OgtF/Y(mER-Cre) MEFs were treated with 100 µ​M 4-thiouridine (4SU) followed by heat stress at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cells were then irradiated with UV at 365 nm and cell lysates were immunoprecipitated using an anti-eIF4GI antibody. Immunoprecipitates were treated with proteinase K followed by RNA extraction and RT-qPCR. Basal levels of both transcripts in wild-type cells (–HS, –4HT) are set to 1. Error bars, mean ±​ s.e.m.; n =​ 3 independent experiments. d, MEF cells were transfected with plasmids expressing YFP-flag-tagged eIF4GI wild type (WT) or mutant (S68A) for 24 h. Transfected cells were treated with 100 µ​M 4SU followed by heat stress at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cells were then irradiated at 365 nm UV and cell lysates were immunoprecipitated using an anti-Flag antibody. Immunoprecipitates were treated with proteinase K followed by RNA extraction and RT-qPCR. YFP was used as a negative control. Basal levels of both transcripts in wild-type cells (–HS, –4HT) are set to 1. Error bars, mean ±​ s.e.m.; n =​ 3 independent experiments. was essential for SG assembly24. To directly address this issue, we conditions18–20. Having observed defective Hsp70 synthesis and the assessed SG formation in MEF cells with or without OGT knockout marked mRNA trapping by eIF4GI in the absence of O-GlcNAc after exposure to sodium arsenite, a potent SG inducer used in the modification, we reasoned that these stress mRNAs were likely to previous study. Sodium arsenite treatment readily triggered SG for- be retained within SGs. To test this possibility, we used fluorescence mation in both wild-type and OGT knockout cells (Supplementary in situ hybridization (FISH) to detect the cellular location of Hspa1a Fig. 5a). In addition, HeLa cells with OGT knockdown also exhib- induced by heat shock stress16. In wild-type MEFs, FISH signals ited a comparable number of, if not more, SG puncta upon sodium distributed uniformly in the cytoplasm, as well as nuclear foci that arsenite exposure (Supplementary Fig. 5b). We conclude that OGT accounted for active Hsp70 transcription (Fig. 4c). In MEFs lacking is dispensable for SG formation. In the case of heat shock stress, OGT, however, the cytoplasmic FISH signals formed clear puncta O-GlcNAcylation of eIF4GI appears to be essential in SG disas- overlapping with SG markers such as G3BP1. Therefore, in the sembly, presumably by promoting selective translation of Hsp7033. absence of O-GlcNAcylation, Hspa1a mRNA is largely entrapped Indeed, adding back Hsp70 into OGT knockout cells via recombi- within SGs. nant adenoviruses prevented the sustained SG formation induced To confirm that it is the modification of eIF4GI that con- by heat stress (Supplementary Fig. 6). tributes to the mRNA triage in SG dynamics, we compared SG induction in MEF cells transfected with YFP-flag-eIF4GI eIF4GI O-GlcNAcylation releases stress mRNAs from SGs. Stress or the S68A mutant. In spite of the presence of endogenous mRNAs such as Hspa1a are largely excluded from SG recruit- eIF4GI, overexpression of the nonmodifiable S68A mutant was ment, likely because of their preferential translation under stress sufficient for sustained SG formation in response to heat shock

912 Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology NAtURE CHEMIcAl BIOlOGy Articles

a Insol Sol Total b Ogt F /Y(mER-Cre) MEFs 4HT –– ++–– ++ –– ++ – 4HT HS –+ ––++––+ + – + eIF4E eIF4GI Merge + 4HT 8 eIF4GI ** – 4HT 6 PABP1 4 eIF4E

SG index (au) 2 eIF2α + 4HT 0 0 h2 h OGT Post-HS

cdF /Y Ogt (mER-Cre) MEFs YFP-flag-eIF4GI WT G3BP1 Hspa1a Merge YFP eIF4E Merge S68A 8

6 ** – 4HT WT

4

SG index (au) 2 + 4HT S68A 0 0 h2 h Post-HS

Fig. 4 | Depletion of O-GlcNAcylation from eIF4GI prevents SG disassembly. a, OgtF/Y(mER-Cre) MEFs were heat stressed (HS) at 42 °C for 1 h and recovered at 37 °C for 2 h. Whole-cell lysates were fractionated into soluble (Sol; supernatant) and insoluble (Insol; pellet) fractions by centrifugation at 16,000 g for 15 min at 4 °C. Both fractions and total samples were collected for immunoblotting. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. b, OgtF/Y(mER-Cre) MEFs were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2 h. Cells were fixed with 3.7% formaldehyde followed by immunofluorescence staining. Scale bars, 10 µ​m. Right: SGs were quantified by defining an index (SG area/cell area) based on eIF4GI immunofluorescence. Data are presented as arbitrary units (au). Error bars indicate mean ±​ s.e.m. for at least three independent experiments in which four fields of view with at least 30 cells per field of view were quantified. **P <​ 0.01; two-way ANOVA. c, OgtF/Y(mER-Cre) MEFs were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cells were then fixed with 3.7% formaldehyde followed by simultaneous G3BP1 immunofluorescence staining and Hspa1a FISH. Scale bars, 10 µ​m. Representative cells are shown from among 120 cells. d, MEF cells were transfected with plasmids expressing YFP-flag-tagged eIF4GI wild type (WT) or mutant (S68A) for 24 h. Transfected cells were heat stressed at 42 °C for 1 h and recovered at 37 °C for 4 h. Cells were fixed with 3.7% formaldehyde followed by immunofluorescence staining. Scale bars, 10 µ​m. Right: SG index based on YFP. Error bars indicate mean ±​ s.e.m. for at least three independent experiments in which 120 YFP-positive cells were quantified. **P <​ 0.01; two-way ANOVA. stress (Fig. 4d). Additionally, the cytoplasmic puncta formed by Relying on a single guide RNA (sgRNA) targeting Ser67 and the eIF4GI mutant contained strong FISH signals for Hspa1a Ser68 (Supplementary Fig. 7b), we established several MEF cell lines (Supplementary Fig. 5c). These results demonstrate the domi- harboring different Eif4g1 mutations. As expected, the majority of nant-negative feature of eIF4GI mutants lacking O-GlcNAc modi- mutants exhibited an insert/deletion (indel) that resulted in frame- fication. It also suggests that stress-induced O-GlcNAcylation of shift of Eif4g1. For instance, in a cell line bearing a single A insertion eIF4GI tends to liberate stress mRNAs from SGs, thereby permit- before the Ser67 codon, very little full-length eIF4GI was detectable ting selective translation. (Fig. 5b). Intriguingly, the short isoforms of eIF4GI became more abundant in these cells, presumably as a result of cellular adaptation. eIF4GI O-GlcNAcylation controls SG dynamics via PABP1. OGT Since the physiological eIF4GI short isoforms lack the N-terminal targets hundreds of cellular proteins for O-GlcNAcylation1,2. It is region for O-GlcNAc modification, they offers a valuable tool for unclear whether the modification status of eIF4GI alone is suffi- assessing the functionality of eIF4GI lacking both O-GlcNAc modifi- cient to influence SG dynamics, as well as selective mRNA transla- cation and PABP1 binding. Despite the absence of full-length eIF4GI, tion in response to heat stress. To address this question, we used the induction of Hsp70 synthesis upon heat shock stress remained the CRISPR/Cas9 system to genetically engineer the endogenous robust in these cells (Fig. 5b). However, these eIF4GI short isoforms Eif4g1. In mammals, the Eif4g1 gene gives rise to five documented neither changed their solubility nor supported SG formation upon isoforms via alternative promoters, splice sites and start codons34. heat shock stress (Fig. 5d,e). Thus, eIF4GI–PABP1 interaction is cru- These eIF4GI variants can be resolved by prolonged gel electro- cial for SG formation but dispensable for selective Hsp70 synthesis. phoresis (Fig. 5a). O-GlcNAc enrichment from heat-stressed cells To confirm the critical role of PABP1 in SG induction, we assessed followed by eIF4GI blotting revealed only the largest two isoforms SG formation in MEF cells with PABP1 knockdown. Remarkably, (Fig. 5a, right), further suggesting that it is the N-terminal region of even partial depletion of PABP1 nearly abolished SG assembly upon eIF4GI that undergoes O-GlcNAcylation in response to heat shock heat shock stress (Supplementary Fig. 8a). However, there was little stress. OGT has been reported to mediate site-specific protein effect on stress-induced Hsp70 synthesis (Supplementary Fig. 8b). cleavage, as exemplified by host cell factor-1 (HCF-1)35. In MEFs The PABP1-independent Hsp70 translation is in agreement with lacking OGT, however, the abundance of eIF4GI isoforms remained the notion that the selective translation of stress mRNAs occurs unchanged (Supplementary Fig. 7a). preferentially during heat stress recovery.

Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology 913 Articles NAtURE CHEMIcAl BIOlOGy

a S68 b eIF4GI WT Ins(A) PABP O-GlcNAc eIF4GI eIF4GI N Post-HSPN ost-HS e (197) Time 0 01224466001 (h) IP Post-IP IP Post-IP d (165) a c (88) a eIF4GI b (41) eIF4GI e a (1) e Hsp70 c Hsp90 eIF4GI Ins(A) ΔSA ΔSSA N Post-HS N Post-HS N Post-HS Time 02446600224 6(h) a eIF4GI e MEF e Ins(A) Hsp70 eIF4E eIF4GI Merge ΔSSA 8 ** Hsp90 ** Ins(A) 6 d + HS eIF4GI Ins(A) ΔSSA Insol Sol Total Insol Sol Total 4 HS –+––++––++– + a SG index (au) 2 eIF4GI e ΔSSA + HS 0 Hsp90 0 h2 h Post-HS

Fig. 5 | The N terminus of eIF4GI acts a functional switch via O-GlcNAcylation. a, Left: N-terminal region of eIF4GI showing possible isoforms, with the number in parentheses indicating the position of the first amino acid. Both the O-GlcNAc modification site S68 and the PABP1 binding domain are highlighted. Right: eIF4GI isoforms obtained from heat-stressed MEF cells before and after immunoprecipitation (IP) using either O-GlcNAc or eIF4GI antibodies. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. b, Genomically engineered MEF cells bearing wild type (WT) or A insertion (Ins(A)) mutant eIF4GI were heat stressed (HS) at 42 °C for 1 h and recovered at 37 °C for indicated times. Whole-cell lysates were collected for immunoblotting. Representative results are shown from three independent experiments. Uncropped blots are shown in Supplementary Fig. 9. N: no heat shock. c, Genomically engineered MEF cells bearing the eIF4GI Ins(A) mutant or a deletion mutant (Δ​SA and Δ​SSA) were heat stressed at 42 °C for 1 h and recovered at 37 °C for indicated times. Whole-cell lysates were collected for immunoblotting. Uncropped blots are shown in Supplementary Fig. 9. d, Genomically engineered MEF cells bearing the eIF4GI Ins(A) mutant or a deletion mutant (ΔSSA)​ were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2 h. Whole-cell lysates were fractionated into soluble (Sol; supernatant) and insoluble (Insol; pellet) fractions by centrifugation at 16,000 g for 15 min at 4 °C. Both fractions and total samples were collected for immunoblotting. Uncropped blots are shown in Supplementary Fig. 9. e, Genomically engineered MEF cells bearing the eIF4GI Ins(A) mutant or a deletion mutant (ΔSSA)​ were heat stressed at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Cells were fixed with 3.7% formaldehyde followed by immunofluorescence staining. Scale bars, 10 µ​m. Right: SG index based on eIF4GI immunostaining. Error bars indicate mean ±​ s.e.m. for at least three independent experiments in which four fields of view with at least 30 cells per field of view were quantified. **P <​ 0.01; two-way ANOVA.

We next attempted to establish MEF cell lines with genomic in- heat shock stress (Supplementary Fig. 8d). This result indicates frame Eif4g1 mutations in order to generate full-length eIF4GI that that other eIF4G homologs, such as eIF4GII and DAP536, could maintains the PABP1 binding capacity but cannot be modified by not compensate for the eIF4GI loss of function in selective trans- O-GlcNAc. Among hundreds of cell clones created by CRISPR/ lation of Hspa1a mRNA. Notably, lack of eIF4GI completely pre- Cas9, two eIF4GI mutants lacking Ser68 (Δ​SA and Δ​SSA) were vented heat-stress-induced SG formation (Supplementary Fig. 8e). successfully selected (Supplementary Fig. 7b). Unlike the frame- We then reintroduced different variants of eIF4GI, including the shifting Ins(A) mutant, the in-frame deletion mutants showed pre- shortest isoform, M197, which lacks the PABP1 binding site and a dominantly full-length eIF4GI (Fig. 5c). As expected, both Δ​SSA truncated variant, M457, which lacks the entire N-terminal region and Ins(A) mutants showed no discernable O-GlcNAc signals upon (Fig. 6a). To prevent downstream initiation of these transgenes, we heat shock stress (Supplementary Fig. 8c). With the intact PABP1 blocked their N termini by fusing a YFP module. We also intro- interaction, the eIF4GI Δ​SSA mutant quickly lost solubility upon duced synonymous mutations into these constructs rendering them heat shock stress and readily formed SG puncta (Fig. 5d,e). More resistant to sgRNA targeting P700. All the constructs except the critically, stress-induced Hsp70 synthesis was severely impaired in S68A mutant restored stress-induced Hsp70 synthesis (Fig. 6a). these cells (Fig. 5c). Therefore, SG disassembly and selective mRNA However, adding back M197 or M457 mutants failed to trigger SG translation proceed in a coordinated manner during stress recovery induction upon heat shock stress (Fig. 6b). Despite the relatively low from heat shock. expression levels in transfected cells, both M197 and M457 mutants rescued Hsp70 synthesis as fully as the wild type. Therefore, an O-GlcNAcylation acts as a functional switch for eIF4GI. To sub- intact PABP1 binding site is not essential for selective translation stantiate the finding that the O-GlcNAcylation of eIF4GI acts as a of Hsp70. In fact, a persistent PABP1 association (as in the case of functional switch between SG induction and selective mRNA trans- the S68A mutant) prevents subsequent stress mRNA translation. lation, we conducted rescue experiments. To deplete endogenous Further supporting the crucial role of O-GlcNAcylation in the eIF4GI variants, we introduced indel mutations into Eif4g1 by tar- functional switch of eIF4GI, only the wild-type eIF4GI was able to geting the conserved downstream region (P700) using CRISPR/Cas9 ­rescue both SG formation and Hsp70 synthesis (Fig. 6a,b). (Supplementary Fig. 7b). All surviving cells still contained residual A model emerges from our data: the dynamic interaction eIF4GI but failed to induce robust Hsp70 synthesis in response to between eIF4GI and PABP1 serves as a regulatory key to SG

914 Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology NAtURE CHEMIcAl BIOlOGy Articles

abS68 YFP G3BP1 Merge c P700 Hsp70 SG WT 4G 4G S68A WT M197 HS M457

sgRNA Scram 4GI(sgP700) S68A NS Post-HS Stress OGT recovery YFP +++ – – – – Hsp mRNA WT – ––+ – – – YFP- S68A – – – – + – – Flag- M197 – – – – ––+ 4GI M457 ––– – –– + M197

Hsp70

WT/S68A M197 4G 4G Anti-Flag M457 * M457 Stress mRNA General mRNA * translation translation

Fig. 6 | Rescuing effects of eIF4GI variants in SG dynamics and Hsp70 synthesis. a, Genomically engineered MEF cells with eIF4GI knockdown mediated by sgRNA targeting P700 were transfected with plasmids expressing eIF4GI wild type (WT), S68A mutants or short isoforms (M197 or M457). Cells were heat stressed (HS) 24 h after transfection at 42 °C for 1 h and recovered at 37 °C for 2.5 h. Top: eIF4G variants and summary of their rescuing effects in Hsp70 synthesis and SG formation. Bottom: immunoblotting results using whole-cell lysates from transfected cells with heat stress and 6 h recovery. Representative results are shown from three independent experiments. *, nonspecific bands. Uncropped blots are shown in Supplementary Fig. 9. b, The same cells as used in a with G3BP1 staining immediately after heat shock treatment. Scale bars, 10 µm.​ Representative cells are shown from among 120 cells. c, A working model depicts the role of eIF4GI O-GlcNAcylation in functional switch between SG formation and selective translation. Persistent eIF4GI and PABP1 association promotes cap-dependent mRNA translation under the normal growth condition while triggering SG induction upon heat shock stress. Stress-induced O-GlcNAcylation of eIF4GI displaces PABP1, thereby dissolving SGs and permitting selective translation of stress mRNAs.

­dynamics, as well as selective mRNA translation. Persistent eIF4GI O-GlcNAcylation of eIF4GI disrupts the closed mRNA loop. As a and PABP1 association promotes cap-dependent mRNA translation result, global protein synthesis remains repressed, whereas selective under the normal growth condition while triggering SG induction translation of stress mRNAs is permitted. Intriguingly, a wide range upon heat-stress-induced translation arrest (Fig. 6c). Stress-induced of viruses frequently target eIF4GI and PABP1, thereby subverting O-GlcNAcylation of eIF4GI dissociates PABP1, thereby dissolving the host translation machinery38,39. For instance, the picornaviral 2 A SGs followed by selective translation of stress mRNAs. Our data protease cleaves eIF4G40, whereas rotaviral NSP3 disrupts PABP1 reveal a functional switch of eIF4GI that is controlled via reversible and eIF4GI interaction41. Unlike the virus-mediated remodeling of O-GlcNAcylation at the N terminus. By controlling direct PABP1 eIF4F, which is long-lasting, stress-induced O-GlcNAc modifica- interaction, this dynamic modification of eIF4GI coordinates SG tion of eIF4GI is dynamic and reversible. This physiological switch induction and selective translation of stress mRNAs. enables cells to produce an adaptive translatome by turning on differ- ent modes of translation initiation in response to stress. Discussion Perhaps the most surprising finding is that depleting O-GlcNAc Environmental stress is detrimental to cell viability and requires an modification or genetically eliminating the modification site of adequate reprogramming of cellular activities to maximize cell sur- eIF4GI leads to mRNA trapping in the insoluble fraction. This para- vival. Proteins that confer adaptive benefits are often preferentially doxical result is in accordance with the sustained SG formation in synthesized; this is subject to extensive regulation at different levels. the absence of eIF4GI O-GlcNAcylation. SG formation is believed to Dynamic O-GlcNAc modification has been shown to integrate envi- be crucial for stress adaptation and is tightly coupled with the pres- ronmental cues into a wide range of fundamental cellular processes, ence of mRNAs stalled in translation initiation18–21. Although both including transcription, protein degradation, basal metabolism and eIF4GI and PABP1 are core SG markers, it is intriguing to find that signaling pathways1,2. Our finding that dynamic O-GlcNAc modifi- their interaction contributes to SG dynamics, at least during heat cation acts as a functional switch for eIF4GI imparts an additional shock stress. A recent study reported that PABP1 in yeast (Pab1) layer of regulation in maintaining protein homeostasis. mediates liquid–liquid phase separation in vivo and in vitro upon The initiation factor eIF4GI acts as a major hub in translational elevated temperatures42. However, SGs appear to form through the control of many cellular processes, stress response and viral infec- crosslinking of nontranslating mRNAs, which provide a scaffold tion37. The N-terminal region of eIF4GI, although largely unstruc- or platform for multiple RNA-binding proteins21. Our findings tured, contains the PABP1 and eIF4E binding sites. PABP1 is known echo the importance of mRNA conformation in SG assembly and to enhance cap- and poly(A)-dependent mRNA translation via direct disassembly. We propose that persistent eIF4GI–PABP1 interac- interaction with eIF4GI29,30. In heat-stressed cells, however, PABP1 is tion, when it is not engaged for active translation, readily poises rapidly dissociated from eIF4GI. This feature is in line with the finding the looped mRNA for SG formation in the cytoplasm. Disrupting that selective translation of Hsp70 mRNA is neither cap- nor PABP1- eIF4GI–PABP1 association by either O-GlcNAc modification or dependent15. However, very little is known how PABP1 is evicted deleting the N-terminal region of eIF4GI helps dissolve SGs. The from the eIF4F complex during heat stress. We provide evidence identification of a single modification that toggles eIF4GI between that N-terminal O-GlcNAcylation of eIF4GI operates to remodel the inhibitors and drivers of SG assembly reshapes our current percep- initiation complex eIF4F. By displacing PABP1, the stress-induced tions of SG regulation. Collectively, dynamic O-GlcNAcylation of

Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology 915 Articles NAtURE CHEMIcAl BIOlOGy eIF4GI serves as a common mechanism underlying selective mRNA 25. Gradi, A. et al. A novel functional human eukaryotic translation initiation translation and SG dynamics. Since aberrant eIF4GI expression factor 4G. Mol. Cell. Biol. 18, 334–342 (1998). 26. Wang, Z. et al. Extensive crosstalk between O-GlcNAcylation and and dysregulated SG formation are associated with certain types of phosphorylation regulates cytokinesis. Sci. Signal. 3, ra2 (2010). 43–45 cancer and neurodegeneration , our findings raise the possibil- 27. Sonenberg, N. & Dever, T. E. Eukaryotic translation initiation factors and ity that new strategies for converting functional variants of eIF4GI regulators. Curr. Opin. Struct. Biol. 13, 56–63 (2003). could be effective for treating these diseases. 28. Howard, A. & Rogers, A. N. Role of translation initiation factor 4G in lifespan regulation and age-related health. Ageing Res. Rev. 13, 115–124 (2014). 29. Gallie, D. R. Te cap and poly(A) tail function synergistically to regulate Methods mRNA translational efciency. Genes Dev. 5, 2108–2116 (1991). Methods, including statements of data availability and any asso- 30. Preiss, T. & Hentze, M. W. Dual function of the messenger RNA cap ciated accession codes and references, are available at https://doi. structure in poly(A)-tail-promoted translation in yeast. Nature 392, org/10.1038/s41589-018-0120-6. 516–520 (1998). 31. Yanagiya, A. et al. Requirement of RNA binding of mammalian eukaryotic Received: 4 December 2017; Accepted: 10 July 2018; translation initiation factor 4GI (eIF4GI) for efcient interaction of eIF4E with the mRNA cap. Mol. Cell. Biol. 29, 1661–1669 (2009). Published online: 20 August 2018 32. Jain, S. et al. ATPase-modulated stress granules contain a diverse proteome and substructure. Cell 164, 487–498 (2016). References 33. Walters, R. W., Muhlrad, D., Garcia, J. & Parker, R. Diferential efects of Ydj1 1. Hardivillé, S. & Hart, G. W. Nutrient regulation of signaling, transcription, and Sis1 on Hsp70-mediated clearance of stress granules in Saccharomyces and cell physiology by O-GlcNAcylation. Cell Metab. 20, 208–213 (2014). cerevisiae. RNA 21, 1660–1671 (2015). 2. Bond, M. R. & Hanover, J. A. A little sugar goes a long way: the cell biology 34. Coldwell, M. J. & Morley, S. J. Specifc isoforms of translation initiation of O-GlcNAc. J. Cell Biol. 208, (869–880 (2015). factor 4GI show diferences in translational activity. Mol. Cell. Biol. 26, 3. Yang, X. & Qian, K. Protein O-GlcNAcylation: emerging mechanisms and 8448–8460 (2006). functions. Nat. Rev. Mol. Cell Biol. 18, 452–465 (2017). 35. Capotosti, F. et al. O-GlcNAc transferase catalyzes site-specifc proteolysis of 4. Janetzko, J. & Walker, S. Te making of a sweet modifcation: structure and HCF-1. Cell 144, 376–388 (2011). function of O-GlcNAc transferase. J. Biol. Chem. 289, 34424–34432 (2014). 36. Liberman, N. et al. DAP5 associates with eIF2β​ and eIF4AI to promote 5. Keembiyehetty, C. et al. Conditional knock-out reveals a requirement for Internal Ribosome Entry Site driven translation. Nucleic Acids Res 43, O-linked N-acetylglucosaminase (O-GlcNAcase) in metabolic homeostasis. 3764–3775 (2015). J. Biol. Chem. 290, 7097–7113 (2015). 37. Jackson, R. J., Hellen, C. U. & Pestova, T. V. Te mechanism of eukaryotic 6. Radermacher, P. T. et al. O-GlcNAc reports ambient temperature and confers translation initiation and principles of its regulation. Nat. Rev. Mol. Cell Biol. heat resistance on ectotherm development. Proc. Natl Acad. Sci. USA 111, 11, 113–127 (2010). 5592–5597 (2014). 38. Walsh, D. & Mohr, I. Viral subversion of the host protein synthesis 7. Groves, J. A., Lee, A., Yildirir, G. & Zachara, N. E. Dynamic O-GlcNAcylation machinery. Nat. Rev. Microbiol. 9, 860–875 (2011). and its roles in the cellular stress response and homeostasis. Cell Stress 39. McCormick, C. & Khaperskyy, D. A. Translation inhibition and stress Chaperones 18, 535–558 (2013). granules in the antiviral immune response. Nat. Rev. Immunol. 17, 8. Zachara, N. E. et al. Dynamic O-GlcNAc modifcation of nucleocytoplasmic 647–660 (2017). proteins in response to stress. A survival response of mammalian cells. J. Biol. 40. Schneider, R. J. & Mohr, I. Translation initiation and viral tricks. Chem. 279, 30133–30142 (2004). Trends Biochem. Sci. 28, 130–136 (2003). 9. Martinez, M. R., Dias, T. B., Natov, P. S. & Zachara, N. E. Stress-induced 41. Piron, M., Vende, P., Cohen, J. & Poncet, D. Rotavirus RNA-binding protein O-GlcNAcylation: an adaptive process of injured cells. Biochem. Soc. Trans. NSP3 interacts with eIF4GI and evicts the poly(A) binding protein from 45, 237–249 (2017). eIF4F. EMBO J 17, 5811–5821 (1998). 10. Kazemi, Z., Chang, H., Haserodt, S., McKen, C. & Zachara, N. E. O-linked 42. Riback, J. A. et al. Stress-triggered phase separation is an adaptive, β​-N-acetylglucosamine (O-GlcNAc) regulates stress-induced heat shock evolutionarily tuned response. Cell 168, 1028–1040.e1019 (2017). protein expression in a GSK-3β​-dependent manner. J. Biol. Chem. 285, 43. Silvera, D., Formenti, S. C. & Schneider, R. J. Translational control in cancer. 39096–39107 (2010). Nat. Rev. Cancer 10, 254–266 (2010). 11. Richter, K., Haslbeck, M. & Buchner, J. Te heat shock response: life on the 44. Grabocka, E. & Bar-Sagi, D. Mutant KRAS enhances tumor cell ftness by verge of death. Mol. Cell 40, 253–266 (2010). upregulating stress granules. Cell 167, 1803–1813.e1812 (2016). 12. Bukau, B., Weissman, J. & Horwich, A. Molecular chaperones and protein 45. Truitt, M. L. & Ruggero, D. New frontiers in translational control of the quality control. Cell 125, 443–451 (2006). cancer genome. Nat. Rev. Cancer 16, 288–304 (2016). 13. Morimoto, R. I. Regulation of the heat shock transcriptional response: cross talk between a family of heat shock factors, molecular chaperones, and Acknowledgements negative regulators. Genes Dev. 12, 3788–3796 (1998). We thank N. E. Zachara for providing inducible MEF Ogt knockout cell lines. We also 14. Anckar, J. & Sistonen, L. Regulation of HSF1 function in the heat stress thank Qian lab members for discussions. We are grateful to Cornell University Life response: implications in aging and disease. Annu. Rev. Biochem. 80, Sciences Core Laboratory Center for confocal microscope imaging support and mass 1089–1115 (2011). spectrometry. We thank the Proteomic and MS Facility of Cornell University for 6 15. Zhou, J. et al. Dynamic m A mRNA methylation directs translational control providing the mass spectrometry data. The Orbitrap Fusion mass spectrometer was of heat shock response. Nature 526, 591–594 (2015). supported by US National Institutes of Health SIG grant 1S10 OD017992-01. This 16. Vera, M. et al. Te translation elongation factor eEF1A1 couples transcription work was supported by grants to S.-B.Q. from the US National Institutes of Health to translation during heat shock response. Elife 3, e03164 (2014). (R01AG042400 and R01GM1222814) and a HHMI Faculty Scholar grant (55108556). 17. Zander, G. et al. mRNA quality control is bypassed for immediate export of stress-responsive transcripts. Nature 540, 593–596 (2016). Author contributions 18. Buchan, J. R. & Parker, R. Eukaryotic stress granules: the ins and outs of X.Z. and S.-B.Q conceived the project and designed the experiments. X.Z. performed the translation. Mol. Cell 36, 932–941 (2009). majority of experiments. X.E.S. assisted the experiments of eIF4GI isoforms and different 19. Anderson, P. & Kedersha, N. Stress granules: the Tao of RNA triage. chaperones in OGT knockout cells. S.-B.Q wrote the manuscripts. All authors discussed Trends Biochem. Sci. 33, 141–150 (2008). the results and edited the manuscript. 20. Cherkasov, V. et al. Coordination of translational control and protein homeostasis during severe heat stress. Curr. Biol. 23, 2452–2462 (2013). 21. Protter, D. S. & Parker, R. Principles and properties of stress granules. Competing interests Trends Cell Biol 26, 668–679 (2016). The authors declare no competing interests. 22. Sun, J., Conn, C. S., Han, Y., Yeung, V. & Qian, S. B. PI3K-mTORC1 attenuates stress response by inhibiting cap-independent Hsp70 translation. Additional information J. Biol. Chem. 286, 6791–6800 (2011). Supplementary information is available for this paper at https://doi.org/10.1038/ 23. Zeidan, Q., Wang, Z., De Maio, A. & Hart, G. W. O-GlcNAc cycling enzymes s41589-018-0120-6. associate with the translational machinery and modify core ribosomal Reprints and permissions information is available at www.nature.com/reprints. proteins. Mol. Biol. Cell 21, 1922–1936 (2010). 24. Ohn, T., Kedersha, N., Hickman, T., Tisdale, S. & Anderson, P. A functional Correspondence and requests for materials should be addressed to S.-B.Q. RNAi screen links O-GlcNAc modifcation of ribosomal proteins to stress Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in granule and processing body assembly. Nat. Cell Biol. 10, 1224–1231 (2008). published maps and institutional affiliations.

916 Nature Chemical Biology | VOL 14 | OCTOBER 2018 | 909–916 | www.nature.com/naturechemicalbiology NAtURE CHEMIcAl BIOlOGy Articles

Methods Higher Collision dissociation product ion triggered ETD (HCD-pd-ETD) MS/MS Cell lines and reagents. OgtF/Y(mER-Cre) and OgtF/Y(GFP) MEFs were kindly scans for precursor peptides with 2–8 charges above a threshold ion count of provided by Natasha E. Zachara (Te Johns Hopkins University School of 50,000 with normalized collision energy of 32%. Product ion trigger list consisted Medicine) and were maintained in Dulbecco’s Modifed Eagle’s Medium (DMEM) of peaks at 204.0867 Da (HexNAc oxonium ion) or 138.0545 Da (HexNAc with 10% FBS. For heat shock treatment, MEF cells were incubated at 42 °C fragment). The ETD reaction time was set at 150 ms and the SA energy was set at water bath for 1 h followed by recovery at 37 °C. 4-Hydroxytamoxifen (H7904), 20%. All data were acquired under Xcalibur 3.0 operation software and Orbitrap 4-thiouridine (T4509), sodium arsenite and thiamet G (SML0244) were purchased Fusion Tune Application v. 2.1 (Thermo-Fisher Scientific). from Sigma. Staurosporine (9953) and anti-eIF2α​ (9722) were purchased from All MS and MS/MS raw spectra from each sample were searched using Cell Signaling. Anti-O-GlcNAc (RL2, MA1-072), Alexa Fluor 488- or Alexa Fluor Byonic v. 2.8.2 (Protein Metrics, San Carlos, CA) using a mouse protein database. 546-labeled secondary antibodies (A21202, A10040) and Hoechst were purchased The peptide search parameters were as follow: two missed cleavage for full from Fisher Scientifc; anti-eIF4GI (sc-11373), anti-eIF4E (sc-9976) from Santa trypsin digestion with fixed carbamidomethyl modification of cysteine, variable Cruz; anti-Flag (M2) and β​-actin monoclonal antibody (F1804) from Sigma; anti- modifications of methionine oxidation and deamidation on asparagine or OGT (11576-2-AP) and anti-G3BP1 (13057-2-AP) from Proteintech; anti-Hsp70 glutamine residues. The peptide mass tolerance was 10 ppm and fragment mass (SPA810) anti-Hsp25 (SPA801) from Enzo Life Sciences; anti-eIF4GI for IF (2858), tolerance values for HCD and EThcD spectra were 0.05 Da and 0.6 Da, respectively. anti-HSP90 (4874), anti-RPS6 (2217), anti-PARP (9542), anti-eIF4E (9742) and Both the maximum number of common and rare modifications were set at two. anti-PABP1 (4992) from Cell Signaling; anti-eIF4GI C-terminal and anti-eIF4GII The glycan search was performed against a list of 78 mammalian O-linked glycans. antibodies were kindly provided by Nahum Sonenberg (McGill University)46. Identified peptides were filtered for maximum 2% FDR. siRNAs targeting human OGT and mouse PABP1 were purchased from Santa Cruz. Plasmids and siRNA transfections were performed using Lipofectamine 2000 Immunoprecipitation. Cells were washed twice with PBS and lysed in lysis buffer (Invitrogen) according to the manufacturer’s instruction. A (20 mM Tris-HCl, pH 7.5, 100 mM KCl, 1 mM dithiothreitol, 0.5 mM EDTA, 10% glycerol, 0.5% NP-40). Lysates were spun at 16,000 g for 15 min at 4 °C and Plasmids. eIF4GI cDNA was cloned from total RNAs extracted from HeLa cells supernatants were collected and incubated with primary antibody and agarose by RT-PCR. The amplified eIF4GI coding sequence was inserted into pcDNA3.1/ beads at 4 °C for overnight. Immunoprecipitates were washed four times with lysis myc-his (Invitrogen). For eIF4GI mutants, mutagenesis was performed using buffer A. The washed beads were resuspended in SDS sample buffer (100 mM Tris, the QuikChange II site-directed mutagenesis kit according to the manufacturer’s pH 6.8, 2% SDS, 15% glycerol, 5% β​-mercaptoethanol, 0.1% bromophenol blue), instruction (Stratagene). All plasmids were confirmed by DNA sequencing. boiled for 5 min and analyzed by immunoblotting.

Generation of eIF4GI mutants cell lines by LentiCRISPRv2. LentiCRISPRv2 Fractionation. Cells were washed twice with PBS and lysed in lysis buffer A on ice plasmids targeting eIF4GI and PABP1 were constructed using methods for 15 min. After spin at 16,000 g for 15 min at 4 °C, three volumes of supernatants previously described47,48. Briefly, complementary oligonucleotides containing was mixed with one volume 4×​ SDS sample buffer and pellets were resuspended in the specific sgRNA sequence and overhangs complementary to overhangs 1×​ SDS sample buffer, then boiled for 5 min, resolved by SDS-PAGE and processed generated by BsmBI digestion of LentiCRISPRv2 were annealed to the for immunoblotting. BsmBI digested LentiCRISPRv2 plasmid to generate the functional transfer vector. Undigested LentiCRISPRv2 plasmid lacking a sgRNA sequence RNA isolation and RT-qPCR. Total cellular RNA was extracted using the was used for pseudovirus production as a control (Scramble). The sgRNA TRIzol reagent (Invitrogen) according to the manufacturer’s instruction. sequence for eIF4GI S68 is ACTCGGGAGGCTGCACTGCT (antisense); Single-strand cDNA synthesis was carried out using the high-capacity cDNA for eIF4GI P700, AGCACCACATCAGTGATATG (antisense); for PABP1, reverse transcription kit (Applied Biosystems) followed by standard PCR GCAAGCCAGTACGCATCATG (sense). Lentiviral particles were packaged using reactions. Real-time PCR reactions were performed using Power SYBR Lenti-X 293 T cells (Clontech). Virus-containing supernatants were collected 48 h Green Master Mix (Applied Biosystems) using a LightCycler 480. Gene after transfection and filtered to eliminate cells, and target cells were infected in expression was normalized to β​-actin cDNA levels and calculated as a relative the presence of 8 µ​g mL−1 Polybrene. Cells were selected 48 h later with 2 μ​g mL−1 gene expression using the 2−ΔΔCt method. Oligonucleotide primers are as puromycin. Single-cell suspensions of selected cells were transferred to 96-well follows: mouse Hspa1a: forward 5′​-TGGTGCAGTCCGACATGAAG-3′​ plates by serial dilution. Single cells were then expanded and analyzed by PCR and reverse 5′-GCTGAGAGTCGTTGAAGTAGGC-3​ ′​; mouse amplification of genomic DNA flanking the CRISPR-targeted region. Forward β​2 microglobulin: forward 5′​-CTGCTACGTAACACAGTTCCACCC-3′​ primer for eIF4GI S68: 5′​AAGCTTAAGGAGGCCACTTAGCCATCAG-3′​; reverse and reverse 5′-CATGATGCTTGATCACATGTCTCG-3​ ′​; mouse primer for eIF4GI S68: 5′​TCTAGAACTACACACTAGGTAACCTGAC-3′​; forward β​-actin: forward 5′-TTGCTGACAGGATGCAGAAG-3​ ′​ and reverse primer for eIF4GI P700: 5′​AAGCTTAGTACAAGTCAGGTGTGTTTAAG-3′​; 5′​-ACTCCTGCTTGCTGATCCACAT-3′​. For ribosome profile analysis, reverse primer for eIF4GI P700: 5′ ​ TCTAGACTCGGGGGCCCACGGTTGCTG-3′​. quantification of target genes of each fraction was normalized using the reference The insertion or deletion of eIF4GI was identified by DNA sequencing. firefly luciferase. Primers used are forward 5′​-ATCCGGAAGCGACCAACGCC-3′​ and reverse 5′​-GTCGGGAAGACCTGCCACGC-3′​. Mass spectrometry to map the O-GlcNAcylation sites of eIF4GI. OgtF/Y(mER-Cre) For zero-distance crosslinking, cells were pretreated with 100 μ​M 4SU and then MEFs were treated with heat shock at 42 °C for 1 h followed by recovery at 37 °C treated with heat shock at 42 °C for 1 h, followed by recovery at 37 °C for indicated for 2.5 h. Cell lysates were then collected followed for eIF4GI immunoprecipitation. times. Cells were washed with PBS once and irradiated at 0.15 J cm−2 total energy Immunopurified eIF4GI was submitted to the Proteomics Facility at Cornell of 365-nm UV light in a Stratalinker 2400. Cells were collected with lysis buffer B University Institute of Biotechnology for MS. For in-gel trypsin digestion, the (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM MgCl2, 1 mM DTT, 1% Triton X-100 250-kDa protein bands from an SDS–PAGE gel were cut into ~1 mm cubes and and 40 U/mL RNase OUT) containing Turbo DNase I (Invitrogen) 25 U mL−1, subjected to in-gel digestion followed by extraction of the tryptic peptide. The and then clarified by centrifugation for 10 min at 20,000 g, 4 °C. Supernatants were excised gel pieces were washed consecutively in 200 μ​L distilled water, 100 mM then incubated with primary antibody and agarose beads for 2 h at 4 °C. The beads ammonium bicarbonate (ambic)/acetonitrile (1:1) and acetonitrile. The gel were washed by lysis buffer B twice, by lysis buffer B containing 0.1% SDS and pieces were reduced with 70 μL​ of 10 mM DTT in 100 mM ambic for 1 h at 0.05% sodium deoxycholate twice, and then by lysis buffer B twice. Washed beads 56 °C, alkylated with 100 μ​L of 55 mM iodoacetamide in 100 mM ambic at room were treated with 1.2 μ​g μ​L−1 proteinase K at 55 °C for 30 min, followed by RNA temperature in the dark for 60 min. After wash steps as described above, the gel extraction and RT-qPCR. slices were dried and rehydrated with 50 µ​L trypsin (Promega) in 50 mM ambic, 10% acetonitrile (20 ng/µ​L) at 37 °C for 16 h. The digested peptides were extracted Immunofluorescence staining. Cells grown on glass coverslips were fixed in twice with 70 μ​L of 50% acetonitrile, 5% formic acid and once with 70 μ​L 3.7% formaldehyde and permeabilized with 0.5% Triton X-100 in PBS. After of 90% acetonitrile, 5% formic acid. Extracts from each sample were combined blocking in 2% BSA in PBS 30 min, fixed cells were incubated with the primary and lyophilized. antibody at 4 °C overnight, followed by 1 h incubation at room temperature with For O-linked glycosylation identification, the in-gel tryptic digests were Alexa Fluor-labeled secondary antibodies. Cells were then washed with PBS reconstituted in 20 μ​L of 0.5% formic acid for nanoLC-ESI-MS/MS analysis, which and incubated for 5 min in PBS supplemented with Hoechst to counterstain was carried out using an Orbitrap Fusion Tribrid (Thermo-Fisher Scientific, San the nuclei. After final wash with PBS, coverslips were mounted onto slides and Jose, CA) mass spectrometer equipped with a nanospray Flex Ion Source coupled viewed using a confocal microscope (Zeiss LSM710) at the Cornell University with a Dionex UltiMate3000RSLCnano system (Thermo, Sunnyvale, CA). The Institute of Biotechnology. gel-extracted peptide samples (10 μ​L) were injected onto a PepMap C-18 RP Stress granules were quantified using ImageJ (https://imagej.nih.gov/ij) as nanotrap column (5 µ​m, 100 µ​m i.d. ×​ 20 mm, Dionex) with nanoViper fittings at previously described44. Briefly, images were randomly acquired in 4 different fields 20 µ​L/min flow rate for online desalting and then separated on a PepMap C-18 with at least 30 cells per field. Background intensity was subtracted from each RP nanocolumn (2 µ​m, 75 µ​m ×​ 25 cm) at 35 °C, and eluted in a 90 min gradient of image. The plug-in “Analyze Particles” was used to measure the total area of stress 5% to 38% acetonitrile in 0.1% formic acid at 300 nL/min. The Orbitrap Fusion granules/total cell area. The stress granule index was determined by computing was operated in positive ion mode as described previously49. The Orbitrap full the total stress granule area in relation to the total cell area for each field and then MS survey scan (m/z 400–1,800) was followed by 3 s ‘Top Speed’ data-dependent averaged across all fields.

Nature Chemical Biology | www.nature.com/naturechemicalbiology Articles NAtURE CHEMIcAl BIOlOGy

RNA FISH. RNA FISH was performed as previously described with small Reporting Summary. Further information on experimental design is available in modifications16. The same set of Stellaris FISH probes, single-labeled (quasar570) the Nature Research Reporting Summary linked to this article. oligonucleotides, were synthesized to bind to mouse Hspa1a from Biosearch Technologies. MEF cells grown on glass coverslips were fixed in 3.7% formaldehyde Data availability. All data generated or analyzed during this study are included and permeabilized with 0.5% Triton X-100 in PBS. Cells were then washed in in this published article (and its Supplementary Information files) or are available PBSM 10 min and incubated 60 min at 37 °C in prehybridization solution (20% from the corresponding author on reasonable request. formamide, 2 ×​ SSC, 2 mg/mL BSA and 200 mM vanadyl ribonucleoside complex). After prehybridization, the hybridization buffer (Biosearch Technologies, References SMF-HB1-10) containing probe (125 nM) plus anti-G3BP1 antibody were added 46. Ramirez-Valle, F., Braunstein, S., Zavadil, J., Formenti, S.C. & Schneider, R.J. to cells and incubated in the dark at 37 °C for 16 h. Cells were then incubated with eIF4GI links nutrient sensing by mTOR to cell proliferation and inhibition of wash buffer A (SMF-WA1-60) plus anti-rabbit second antibody in the dark at autophagy. J. Cell Biol. 181, 293–307 (2008). room temperature for 1 h. Wash buffer A was aspirated, cells were washed twice 47. Sanjana, N. E., Shalem, O. & Zhang, F. Improved vectors and genome-wide with wash buffer B (SMF-WB1-20) and then coverslips were mounted onto slides libraries for CRISPR screening. Nat. Methods 11, 783–784 (2014). with mounting medium containing DAPI and viewed using a confocal microscope 48. Shalem, O. et al. Genome-scale CRISPR-Cas9 knockout screening in human (Zeiss LSM710) at the Cornell University Institute of Biotechnology. cells. Science 343, 84–87 (2014). 49. Tomas, C. J., Cleland, T. P., Zhang, S., Gundberg, C. M. & Vashishth, D. Statistical analysis. Results are expressed as mean ±​ s.e.m. Comparisons between Identifcation and characterization of glycation adducts on osteocalcin. groups were made by two-way ANOVA and unpaired two-tailed Student t-test. Anal. Biochem. 525, 46–53 (2017).

Nature Chemical Biology | www.nature.com/naturechemicalbiology nature research | life sciences reporting summary

Corresponding author(s): Shu-Bing Qian

Life Sciences Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form is intended for publication with all accepted life science papers and provides structure for consistency and transparency in reporting. Every life science submission will use this form; some list items might not apply to an individual manuscript, but all fields must be completed for clarity. For further information on the points included in this form, see Reporting Life Sciences Research. For further information on Nature Research policies, including our data availability policy, see Authors & Referees and the Editorial Policy Checklist.

Please do not complete any field with "not applicable" or n/a. Refer to the help text for what text to use if an item is not relevant to your study. For final submission: please carefully check your responses for accuracy; you will not be able to make changes later.

` Experimental design 1. Sample size Describe how sample size was determined. For independent experiments, at least three replicates are conducted. For stress granule (SG) index calculation, 120 YFP positive cells are quantified for each experiment. 2. Data exclusions Describe any data exclusions. No data were excluded 3. Replication Describe the measures taken to verify the reproducibility Yes, at least three biological replicates of the experimental findings. 4. Randomization Describe how samples/organisms/participants were Independent experiments like immunoblotting cannot be randomized. allocated into experimental groups. 5. Blinding Describe whether the investigators were blinded to Independent experiments like immunoblotting cannot be blinded. group allocation during data collection and/or analysis. Note: all in vivo studies must report how sample size was determined and whether blinding and randomization were used.

6. Statistical parameters For all figures and tables that use statistical methods, confirm that the following items are present in relevant figure legends (or in the Methods section if additional space is needed). n/a Confirmed

The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement (animals, litters, cultures, etc.) A description of how samples were collected, noting whether measurements were taken from distinct samples or whether the same sample was measured repeatedly A statement indicating how many times each experiment was replicated The statistical test(s) used and whether they are one- or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section. A description of any assumptions or corrections, such as an adjustment for multiple comparisons Test values indicating whether an effect is present

Provide confidence intervals or give results of significance tests (e.g. P values) as exact values whenever appropriate and with effect sizes noted. November 2017 A clear description of statistics including central tendency (e.g. median, mean) and variation (e.g. standard deviation, interquartile range) Clearly defined error bars in all relevant figure captions (with explicit mention of central tendency and variation)

See the web collection on statistics for biologists for further resources and guidance.

1 ` Software nature research | life sciences reporting summary Policy information about availability of computer code 7. Software Describe the software used to analyze the data in this Microsoft Excel, SigmaPlot, GraphPad Prism, ImageJ study.

For manuscripts utilizing custom algorithms or software that are central to the paper but not yet described in the published literature, software must be made available to editors and reviewers upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). Nature Methods guidance for providing algorithms and software for publication provides further information on this topic.

` Materials and reagents Policy information about availability of materials 8. Materials availability Indicate whether there are restrictions on availability of All materials are readily available unique materials or if these materials are only available for distribution by a third party. 9. Antibodies Describe the antibodies used and how they were validated anti-eIF2 alpha (9722) was purchased from Cell Signaling. Anti-O-GlcNAc (RL2, MA1-072), for use in the system under study (i.e. assay and species). Alexa Fluor 488 or 546 labeled secondary antibodies (A21202, A10040) and Hoechst were purchased from Fisher Scientific; anti-eIF4GI (sc-11373), anti-eIF4E (sc-9976) from Santa Cruz; anti-Flag (M2) and beta-actin monoclonal antibody (F1804) from Sigma; anti-OGT (11576-2-AP) and anti-G3BP1 (13057-2-AP) from Proteintech; anti-Hsp70 (SPA810) anti- Hsp25 (SPA801) from Enzo Life Sciences; anti-eIF4GI for IF (2858) , anti-HSP90 (4874) , anti- RPS6 (2217), anti-PARP (9542), and anti-eIF4E (9742), and anti-PABP1 (4992) from Cell Signaling; anti-eIF4GI C-terminal and anti-eIF4GII antibodies were kindly provided by Dr. Nahum Sonenberg (McGill University). 10. Eukaryotic cell lines a. State the source of each eukaryotic cell line used. MEF/Ogt F/Y(mER-Cre), MEF/Ogt F/Y(GFP), MEF, HeLa

b. Describe the method of cell line authentication used. Cell lines are not yet authenticated

c. Report whether the cell lines were tested for Cell lines are not yet test for mycoplasma contamination mycoplasma contamination. d. If any of the cell lines used are listed in the database No misidentified cell lines were used of commonly misidentified cell lines maintained by ICLAC, provide a scientific rationale for their use.

` Animals and human research participants Policy information about studies involving animals; when reporting animal research, follow the ARRIVE guidelines 11. Description of research animals Provide all relevant details on animals and/or No animal were used animal-derived materials used in the study.

Policy information about studies involving human research participants 12. Description of human research participants Describe the covariate-relevant population The study did not involve any human research participants characteristics of the human research participants. November 2017

2